The elusive meaning of black paintings and red blocks
May 20, 2020
Joe N, a longtime reader, tweeted about the following chart, by the People's Policy Project:
This is a simple column chart containing only two numbers, far exceeded by the count of labels and gridlines.
I look at charts like the lady staring at these Ad Reinhardts:
My artist friends say the black squares are not the same, if you look hard enough.
Here is what I learned after one such seating:
The tiny data labels sitting on the inside top edges of the columns hint that the right block is slightly larger than the left block.
The five labels of the vertical axis serve no purpose, nor the gridlines.
The horizontal axis for time is reversed, with 2019 appearing after 2020 (when read left to right).
The left block has one month while the right block has 12 months. This is further confused by the word "All" which shares the same starting and ending letters as "April".
As far as I can tell, the key message of this chart is that the month of April has the impact of a full year. It's like 12 months of outflows from employment hitting the economy in one month.
My first response is this chart:
Breaking the left block into 12 pieces, and color-coding the April piece brings out the comparison. You can also see that in 2019, the outflows from employment to unemployment were steady month to month.
Next, I want to see what happens if I restored the omitted months of Jan to March, 2020.
The story changes slightly. Now, the chart says that the first four months have already exceeded the full year of 2019.
Since the values hold steady month to month, with the exception of April 2020, I make a monthly view:
You can see the slight nudge-up in March 2020 as well. This draws more attention to the break in pattern.
For time-series data, I prefer to look at line charts:
As I explained in this post about employment statistics (or Chapter 6 of Numbersense (link)), the Bureau of Labor Statistics classifies people into three categories: Employed, Unemployed and Not in Labor Force. Exits from Employed to Unemployed status contribute to unemployment in the U.S. To depict a negative trend, it's often natural to use negative numbers:
You may realize that this data series paints only a partial picture of the health of the labor market. While some people exit the Employed status each month, there are others who re-enter or enter the Employed status. We should really care about net flows.
In all of 2019, there were more entrants than exits, leading to a slightly positive net inflow to the Employed status from Unemployed (blue line). In April 2020, the red line (exits) drags the blue line dramatically.
Of course, even this chart is omitting important information. There are also flows from Employed to and from Not in Labor Force.
Good analysis as always. I like the monthly barchart. But would hv added a cumulative line for 2019 which would then rise to the April bar height right? This pulls us back to the original headline of the story.
Posted by: Antro | May 20, 2020 at 09:20 PM
Antro: Like the concept. Might need some text to explain the cumulation but worth sketching out.
Posted by: Kaiser | May 21, 2020 at 11:20 AM
You have misread People's Policy Project's intention. The author is saying that employer-based insurance is unreliable, not that April 2020 has been a bad month. People's Policy Project is against employer-based healthcare generally (and in favor of universal healthcare).
Posted by: Casey | May 21, 2020 at 01:14 PM
Casey: I'm not sure how the featured comparison makes an argument for or against employer-based insurance. Can you clarify?
Posted by: Kaiser | May 21, 2020 at 04:17 PM